Several neuroimaging studies of patients with OCD have pointed to basal ganglia and the frontal cortical regions being relevant for an understanding of the pathophysiology and therapy of OCD. In a search for the neural substrate underlying the therapeutic action of paroxetine in the therapy of OCD we measured regional glucose metabolism in a PET study of 20 OCD patients before and after at least 3 months of treatment. We used 18-fluoro-deoxyglucose PET-scanning to measure regional cerebral glucose metabolic rate (rCMRglc) in 20 non-depressed patients fulfilling DSM-IV criteria for OCD. Patients were studied before and after 12-20 wk of treatment with the serotonin re-uptake inhibitor paroxetine. Clinical assessment rating with the Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) was performed before the first and after the second study. The PET data was analysed regionally using statistical parametric mapping (SPM-96). A clinical improvement was indicated by a mean decrease of 55% in the Y-BOCS score. There was no difference in global cerebral metabolism before and after treatment whereas a post-treatment reduction in normalized rCMRglc was found in the right caudate nucleus. This finding also showed a significant positive correlation with symptom severity. Our results support hypotheses regarding a malfunction of the cortico-striato-thalamic system in the pathophysiology of OCD and particularly point to the caudate nucleus playing an important role for the therapeutic action of paroxetine in the treatment of OCD.